According to SIMA, AI agents may be very important in video games in the future. It also moves us one step closer to AI which can work effectively with humans to complete tasks in real-world settings as well as in games.
Scalable Instructable Multiworld Agent, or SIMA, is a new artificial intelligence (AI) gaming agent developed by Google DeepMind that can obey natural language commands to carry out tasks in a variety of video game scenarios. It was unveiled on March 13th. In other words, Google’s latest AI is ready to play games with you.
According to SIMA, AI agents may be very important in video games in the future. It also moves us one step closer to AI which can work effectively with humans to complete tasks in real-world settings as well as in games.
This week has been hectic for AI fans, with the introduction of a completely autonomous AI software engineer and an AI robot to assist elders overcome loneliness. What you need to know about SIMA is as follows.
What is SIMA?
AI Research Lab SIMA is described as an AI Agent by Google Deepmind, distinguishing it from AI models such as OpenAI’s ChatGPT and Google Gemini.
AI models are trained on a large dataset and have little ability to function independently. An AI Agent, on the other hand, can digest data and make decisions on its own.
SIMA can be described as a generalist AI agent capable of performing a variety of activities. It’s like having a virtual companion who can understand and obey orders in a variety of virtual locations, from exploring intriguing dungeons to creating opulent castles. It can do tasks and solve problems that are assigned to it.
It might be compared to a digital explorer—a computer program that is essentially extremely intelligent and capable of understanding your desires and assisting in their creation in the virtual world.
What is SIMA’s working?
Because SIMA has been trained to process human language, it “understands” your commands. It therefore knows exactly what you mean when you give it instructions to build a fortress or locate the treasure box.
This AI agent’s ability to learn and adapt is one of its unique qualities. Through its interactions with the user, SIMA does this.
Through experience and continuous improvement, SIMA becomes more intelligent the more you use it. As a result, it can comprehend and respond to user demands more effectively.
It is a significant accomplishment for an AI system to be able to play even one game, given the state of AI progress today. Beyond that, though, SIMA is able to obey commands in a range of gaming environments. This might bring in more beneficial AI agents for use in different settings.
According to Google DeepMind’s most recent study, it is possible to convert sophisticated AI models’ capabilities into “useful, real-world actions through a language interface.”
Google hopes that by doing this, video games can serve as sandboxes for SIMA and other AI agents to learn how AI systems might be more useful.
A look at SIMA’s training
The company’s description on its official blog reads as follows: “We partnered with game developers to train SIMA on a variety of video games. This research marks the first time an agent has demonstrated it can understand a broad range of gaming worlds and follow natural-language instructions to carry out tasks within them, as a human might.”
Google Deepmind worked with eight-game developers to test SIMA on nine distinct video games, including Teardown by Tuxedo Labs and No Man’s Sky by Hello Games, in order to expose the AI agent to various situations.
With each game, SIMA’s portfolio unlocked a new dynamic environment and a variety of abilities for it to pick up, like basic navigation, menu navigation, resource mining, spaceship piloting, etc.
Additionally, they made use of four study environments, one of which was created using the Unity cross-platform gaming engine. AI agents were needed in this setting, dubbed Construction Lab, to construct sculptures out of building blocks.
The purpose of this exercise was to evaluate the agent’s ability to manipulate objects and intuitive grasp of the physical environment.
The researchers claim that through studying various gaming environments, SIMA is able to understand the relationship between language and behaviour during gameplay.
“Our first approach was to record pairs of human players across the games in our portfolio, with one player watching and instructing the other. We also had players play freely, then rewatch what they did and record instructions that would have led to their game actions,” read the official blog.
